SIR – In a recent association study carried out in US patients with Autism Spectrum Disorder (ASD), Ingram et al,1 have detected linkage disequilibrium (LD) with a mutation of HOXA1, a homeotic gene playing a role in eearly brain formation. This finding is noteworthy since a dysfunction of the HOXA1 gene (or its coexpressed partner, HOXB1, in hindbrain) is potentially related to the appearance of several dysmorphic features frequently detected in autistic patients.2 In contrast, in an attempt to replicate these findings in Sicilian (Italian) autistic patients, we could not detect any linkage between mutations of the HOXA1 or HOXB1 genes and autism.
The mutation analysed in HOXA1 is a base transition from ‘A’ to ‘G’ at position 218 of mRNA (A218G). In the HOXB1 gene, the ‘at risk’ allele consists of a 9-base insertion (‘INS’), the other allele is named (‘+’). In order to verify the existence of LD between these mutations and autism, we have used: (i) TDT analysis in a group of 85 complete trios families and (ii) a conventional ‘case–control’ study with an ethnically matched sample as reference population. All patients (77 males, 12 females accounting for a male/female ratio of 6.4, mean age 10.44 years; SD ±5.68 years) are mentally retarded and met DSM-IV criteria for autistic disorder. In addition, patients were assessed by means of the (i) Childhood Autism Rating Scale, (ii) Brunet–Lezine test, (iii) WISC-R test, (iv) Psychoeducational Profile Revised (PEP-R), (v) Griffith's Mental Developmental Scales and (vi) Leiter International Performance Scale. Patients who were excluded from the study include those displaying at least one item from the following checklist: neurological focal signs or seizures, chromosomal abnormalities, FMR1 gene mutation (fragile X syndrome), and other neurological diseases such as phenylketonuria, neurofibromatosis, tuberous sclerosis, encephalopathies because of congenital infections. ‘Sicilian ancestry’ for all patients, and individuals of the reference population, was ascertained for at least two generations by enquiring about the place of birth of their maternal and paternal grandparents. Hardy–Weinberg equilibrium was tested by the χ2 test for goodness of fit. Z test was used to test for genetic heterogeneity between cases and controls. TDT was performed by using the (b−c)2/(b+c) formula3 and χ2 test. In our series, nearly 50% of the subjects display mild phenotypic features, potentially related to dysfunction of HOXA1 and/or HOXB1 genes. HOXA1 and HOXB1 genotyping was carried out essentially as described by Ingram et al.1 The frequency of the ‘G’ allele of the HOXA1 gene in the group of 85 autistic patients is 12%, a slightly lower value than found in our reference population (15.5%; N=132). For the HOXB1 gene, the frequency of the ‘INS’ allele is 20.6 and 22.5% in autistic patients and control group (N=80), respectively. Deviation from Hardy–Weinberg equilibrium was detected for none of the four gene/sample-of-individuals combinations tested (P>0.08). Comparison of the ‘at risk’ alleles for both HOXA1 (allele ‘G’) and HOXB1 (allele ‘INS’) genes, between cases and ethnically matched controls, did not reveal statistically significant differences (all Z values are between −1.017 and 0.422; P>0.31). For TDT analyses, from 34 heterozygote parents, 19 transmissions and 15 nontransmissions were scored for allele ‘G’ of HOXA1. For HOXB1, from 52 heterozygote parents 30 transmissions and 22 nontransmissions were scored for allele ‘INS’. Consistent with the results of the case–control study, the TDT test was also not significant (HOXA1: TDT=0.471 (P=0.493); HOXB1: TDT=1.23 (P=0.27)). Our failure to replicate the results obtained by Ingram et al.1 for the HOXA1 polymorphism in Sicilian families with autism may not be ascribed to sample size. Calculations (performed by the ‘TDT Power Calculator’4—software version PC 1.2.1) suggest that we would have at least a 90% power to detect linkage and association between the HOXA1 polymorphism and autism (at α=0.05). Of course, we cannot exclude that we may have failed to detect the effect of the HOXA1 polymorphism due, for example, to the marked clinical heterogeneity of the disorder that may have caused random or systematic selection of patients. However, the most obvious differences between the two samples concern ethnicity and geography. These differences may imply a potential heterogeneity of pathogenetic factors (genetic and/or environmental) distinguishing the North American from the Sicilian ASD patients. For example, strong evidences for the existence of multilocus heterogeneity in ASD have been provided by the results of various genome-wide screening studies performed in (ethnically and geographically) different groups of patients.5,6,7 Also, environmental exposure to unknown factors, especially during embryonic development, may in principle results in different phenotypic outcomes, even with the same genotype. In other words, for a complex disorder, like autism, each single genetic determinant of the disease does not have an absolute value; rather it assumes pathogenetic significance only in relation to specific genomic and environmental contexts. Association studies can be very powerful in detecting the effect of a gene on a disease if that effect is not too weak; however, because their validation relies on replication studies, at the same time, they bear the intrinsic limitation of being exposed to the effect of allelic and multilocus heterogeneity which, as we know, underly many inherited diseases. During the preparation of this manuscript, other studies8,9,10 reported a lack of association between HOXA1 (A218G) and autism.
This is a preview of subscription content, access via your institution
